Anurag> Looking at /etc/nwam/known_wifi_nets, is the BSSID really necessary?
Anurag> libdladm can connect fine without the BSSID and the list is just
Anurag> getting cluttered with the same ESSID.  Not using the BSSID also
Anurag> simplifies the generate of secure object names.

James> It's a fair question.  At least in the bugs I'm fixing, I'm leaving
James> it alone.  The NWAM design is tragically impaired by CR 6772510, but
James> I'm not sure what the right thing to do is here, and rewriting the
James> NWAM Phase 0 from scratch -- plus the broken parts of libdladm and
James> the drivers -- just wasn't part of the Phase 0.5 project.

Agreed: this was, in retrospect, a bad design decision that we (officially
it was a team decision, but I was the one who pushed for it, so I deserve
the blame) made back when designing the NWAM prototype in Spring 2006.  We
were worried about the "linksys" problem, where zillions of APs come with
a default ESSID that often is not changed when deployed, and thus we chose
to connect via both ESSID and BSSID.  But as described by these two CRs:

6553137 Improve NWAM WiFi mobility
6773627 nwamd should be less interested in BSSID

this has some serious drawbacks that we did not anticipate:

* other OSs (particularly Mac OS X, more noticeable since the introduction
  of the iPhone) only care about ESSID
* roaming on Sun campuses with lots of APs, all using ESSID "sunwifi" but
  with different BSSIDs, doesn't work so well (extrapolate this to the
  general case of roaming in like environments)
* there are DL/driver level bugs w/rt using BSSID

With hindsight, I'd say the problem we avoided is less of an issue than the
problems that our design decision led to.  As such, I would support changing
the logic so that BSSID is not used by default, but only when needed.

-- John

http://blogs.sun.com/jbeck
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to